Characterization of tertiary interactions in a folded protein by NMR

Apr 29, 1992 - Frits Abildgaard,t§ Anne Marie Munk Jorgensen,1 Jens J. Led,*-* 1 Thorkild Christensen,! Ejnar B. Jensen,! Flemming Junker,! and Henri...
0 downloads 0 Views 2MB Size
Biochemistry 1992, 31, 8587-8596

8587

Characterization of Tertiary Interactions in a Folded Protein by NMR Methods: Studies of pH-Induced Structural Changes in Human Growth Hormone? Frits Abildgaard,fg Anne Marie Munk Jargensen,t Jens J. Led,'.: Thorkild Christensen,ll Ejnar B. Jensen) Flemming JunkerJ and Henrik Dalbagell Department of Chemistry, University of Copenhagen, The H.C. 0rsted Institute, Universitetsparken 5. DK-2100 Copenhagen 0, Denmark, and Novo Nordisk AIS, Novo All?, DK-2880 Bagsvaerd, Denmark Received April 29, I992

ABSTRACT: The pH-induced conformational changes in human growth hormone (hGH) have been studied, using a new quantitative N M R approach that combines 13Clabeling of specific backbone carbonyl carbons with a complete spectral analysis of the corresponding 13C resonances. Thus, a complete analysis of the carbonyl resonances of the 26 Leu residues of hGH and their variation with pH provided detailed information about the equilibrium folding processes of the protein, including information about the kinetics of the folding. By combining this information with the pH dependence of readily identifiable 'H resonances, the pH-induced changes observed in the carbonyl carbon spectra can be associated with specific regions in the protein and can be ascribed to a series of localized adjustments in the tertiary structure, brought about by changes in the hydrogen bond interactions or electrostatic interactions between different residues in the globular folded protein. The preexchange lifetimes of these adjustments range from a fraction of a millisecond to a few milliseconds.

Understanding the principles of protein folding has been a challenge of long standing in the area of molecular biology. According to the framework model (Kim & Baldwin, 1982) the protein folding process involves the early formation of hydrogen-bondedsecondary structures. These events are then followed by reorganization to yield tightly packed domains constituting the tertiary structure. The tertiary structure is stabilized by interactions between different elements of the secondary structure. These interactions are defined by only a few residues, and whereas amino acid substitution at other residue positions are allowed, even minor changes at sites directly involved in the tertiary interactions may affect the folding of the protein drastically (Hughson & Baldwin, 1989). Changing the charge distribution in the protein by changing pH generally affects the stability of the protein and may lead to an unfolding of the native state (Nall et al., 1988). Like the other members of the family of growth hormones, the globular structure of human growth hormone (hGH)' has been shown to be sensitive to changes in pH. The pH-induced changes of hGH have been found to be reversible (Aloj & Edelhoch, 1972;Turner et al., 1983),but unlike bovine growth hormone (Holzman et al., 1990) there does not seem to be any indications of a general unfolding of the secondary structure of hGH (Bewley & Li, 1967; Aloj & Edelhoch, 1972). Thus,

the conformational changes can be considereda reorganization of the globular structure imposed by a change in favorable tertiary interactions. With the presence of different stable tertiary structures within virtually the same framework of secondary structures, hGH offers the opportunity to study in detail the stabilizing interactions responsible for the second stage in the folding within the framework model. For the investigation of the conformationalchanges in hGH we here present a new quantitative NMR approach that combines specific isotopic labeling with a complete spectral analysis. Thus, it is shown that specific isotope labeling of the backbone carbonyl carbons of the leucine residues of hGH allows the structural changes to be monitored at 26 different sites distributed throughout the protein, and it is demonstrated that a complete analysis of the pH dependence of the 13C NMR spectrum can provide specific information about the equilibrium folding processes in hGH, including information about the kinetics. Even though the individual signals from the carbonyl carbons of the 26 leucine residues have not yet been assigned, it is possible, in combination with titration data obtained from 'H NMR spectra of hGH, to correlate the structural changes with specific regions of the protein.

This work was supported by the Danish Technical Research Council, Grants 16-3922.H and 16-4679.H, the Ministry of Industry, Grant 85886, and Julie Damms Studiefond. F.A. acknowledgesa predoctoral fellowship from the Danish Natural Science Research Council, Grants 11-6457 and 1 1-8153. * To whom correspondence should be addressed. 1 University of Copenhagen. 5 Present address: Department of Biochemistry, University of Wisconsin-Madison, 420 Henry Mall, Madison, WI 53706. 11 Novo Nordisk A/S. Abbreviations: hGH, human growth hormone; pGH, porcine growth hormone; NMR, nuclear magnetic resonance; FID, free induction decay; TMS, tetramethylsilane; WALTZ, wideband alternating-phase low-power technique for zero residue splitting; NOESY, two-dimensional NOE spectroscopy; NOE, nuclear Overhauser enhancement; TOCSY, total correlation spectroscopy; COSY, two-dimensional correlation spectroscopy; CD, circular dichroism; IE-HPLC, ion-exchangehigh-performance liquid chromatography; ppm, parts per million; SD, standard deviation.

Unlabeled, authentic hGH was prepared by recombinant DNA technology according to Dalbage et al. (1987). Biosynthetic hGH, labeled with 13Cin the carbonyl carbons of the 26 leucine residues, was purified from Escherichia coli strain MC1061 auxotrophic for the amino acid leucine. Fermentation with [ l-13C]leucine(99% Stohler/Kor) as the only amino acid added to the medium, and purification and removal of the amino terminal extension were carried out as described previously (Christensen et al., 1992,1986; Dalbage et al., 1987). As judged by IE-HPLC, the product was more than 99% pure. Samples for proton NMR measurements were prepared by dissolving 20 mg of recombinant hGH in 1.0 mL of DzO, giving a final concentration of 0.9 mM hGH. Samples for 13C N M R were prepared by dissolving 15 mg of 99%

MATERIALS AND METHODS

0006-2960/92/043 1-8587%03.00/0 0 1992 American Chemical Society

8588 Biochemistry, Vol. 31, No. 36, 1992 [l-13C]Leu-hGH in 1.5 mL of D2O (99.996, Norsk Hydro). Lower concentrations were used near the isoelectric pH of 4.9 (Bewley & Li, 1975). The pH was determined at 24 OC by a Radiometer PHM80 meter equipped with a 3-mm glass electrode. Reported values refer to direct meter readings (Kalinichenko, 1976). The solution pH was adjusted with dilute DCl or NaOD, and pH measurements were performed beforeand after data acquisition. Measurementsof pH agreed to within 0.05 unit before and after spectral acquisition. The ionic strength of the samples varied from 0.010 to 0.100 M during the titration process. NMR experiments were performed on a Bruker AM 500 spectrometer, operating at 500.13 MHz for 'H nuclei and at 125.76 MHz for 13Cnuclei. One-dimensional proton spectra were recorded at 293 K and 305 K. Residual HOD was suppressed by saturation prior to data acquisition. Proton chemical shifts are expressed relative to TMS and were measured with respect to internal 1,4-dioxane at 3.74 ppm. The carbon- 13spectra were recorded at 293 K, with a spectral width of 10 000 Hz in 8192 data points. Broadband proton decoupling was applied using a WALTZ-decoupling sequence (Shaka et al., 1983), and a decoupler power of ca. 0.5 W. The number of averaged transients for each spectrum varied from 12 900 to 47 200. Carbon- 13 chemical shifts are referenced relative to external TMS. The pH dependence of the 13CNMR spectra was analyzed by the method of concerted nonlinear least-squares fitting (Abildgaard et al., 1991), Le., a simultaneous fit to all spectra in the titration data set, using the analytical expression for discrete Fourier-transformed NMR spectra (Abildgaard et al., 1988). The spectra were scaled to the same overall integrated intensity prior to the least-squares analysis and weighted according to their relative noise level. Including a possible pH dependence of the chemical shift and the line width, the applied expression takes the form

Abildgaard et al. where 7 2 = kA is the pH independent, first-order rateconstant for the interconversion of A to B. Therefore eq 4 can be rewritten as A v l p = XAAvt12

+ X&VFp + 4 ? r x ~ x ~ ~ ( A l J(5)~ ~ ) ~ k ~

Thus, the measured exchange rate is the rate of exchange between the conformational states at low and high pH, respectively, of the individual carbons within hGH. Consequently, each resonance is fully described by eight parameters: the chemical shift of the two sites, 6~ and 6 ~ the ; apparent pKa; the mean preexchange lifetime, TA; the intensity of the signal, A; the line width in the two sites, Avf12 and Av:/,; and the phase, 4. The pH dependence of the lH NMR chemical shifts was analyzed by a nonlinear least-squares fit of the equation describing a three-step titration = 61x1 + 62x2 + 63x3

+ 64x4

(6) where sobs is the observed chemical shift at any pH, while 81-64 are the chemical shifts of the four species present with molar fractionsx,-x4, respectively. For pKal0004

-

mOOO2-0004

-

0 001- @ 002 f-Joooo-0001

-

u n a

3.0

4.0

5.0

6.0

FIGURE 3: Summary of the apparent titration parameters for the carbonyl resonances given in Table I. The height of the bars corresponds to the magnitude of the pH-induced chemical shift. The pattern codeshows the T A in seconds, as indicated (na = not available). Table 11: Leucine Signals from Table I Arranged in Groups with Identical PKa Valuesu group no. I

signals no.

PKa

1,29394 5 6,7,8,9, 10 11 12,13 14.15 16, 17, 18, 19 20,21 22,23 24.25 26

3.18 3.58 3.80 4.16 4.28 4.53 4.80 5.26 5.54 6.02 6.53

SA

(s)

~~

I1

I11 IV V VI VI1

VI11

IX X XI

0.00 1 3 0.0007 0.0002 0.0027 0.001 0.0027 0.006 0.02

Signal numbers refer to the numbers in Table I.

The distribution of the 26 leucine residues in hGH is indicated in Figure 2. Although the leucine I3C resonances have not yet been assigned to specific residues in the protein, it is clear that the structural changes involve most parts of the globular structure, considering the distribution of the leucine residues throughout the molecule and the general and extensive variation of the carbonyl spectrum with pH. The data in Table I are summarized in Figure 3. It is clear from this figure that the structural transitions occur over a wide pH range. Since the carbonyl carbons, participating in the same structural transition, will have similar PKa values, clusters are expected to be seen in the diagram. Taking the uncertainty of individual PKa values into account, several groups of signals are found. For a signal to be a member of a certain group, the criterion has been used that the observed PKa value must fall within 1 SD from the weighted average of the PKa value of the group. The resulting groupings of the signals are shown in Table 11. Four resonances form a group (group I) that participates in a structural change with an apparent PKa value of 3.18. As

found for both resonances 1 and 4, the preexchange lifetime, TA, for the carbonyl carbons of the protonated form that are involved in this transition is on the order of 1.3 ms corresponding to a rate constant of k A = 770 s-I. Five resonances form group I11 with a PKa value of 3.80. The rate constant associated with this transition is too fast to be detected. The preexchange lifetimedetermined for resonance 7 is considered to be an artifact, due to the very small value of AVAB(14 Hz) and the considerable crowding in this region of the spectrum. Four resonances form a distinct group, VII, with an average PKa value of 4.80 and a 7 A Close to 2.7 ms ( k A = 370 s-'). As for resonance 19, it is not clear whether this resonance is actually a member of this group. The large uncertainties of the spectral parameters describing resonance 19 indicate that it is subjected to more than one structural transition. Among the remaining signals no clear grouping can be found, suggesting that the observed spectral changes are either due to highly localized perturbations or are the effective, average values of two or more structural transitions. For the signals 24-26, only insignificant pH-induced shifts were observed. Titration of Specific Proton Resonances. In order to correlate the pH-induced changes observed in the carbonyl carbon spectra with specific regions in the protein, the pH dependence of readily identifiable IH NMR signals in hGH were investigated. Two sets of titrations were performed (293 and 305 K, respectively). The IH NMR spectrum of hGH in D20 at 305 K is shown in Figure 4, and the pH dependence of the chemical shifts at 293 K of the well-resolved resonances marked A-L are presented in Figure 5. Up to three significant, apparent titrations were fitted by using eqs 6 and 7. In cases where it would obviously lead to improvements in the fit, Hill coefficientswere fitted as well. The parameters derived from the fitting of the data at 293 and 305 K are given in Tables I11 and IV. The resonances A-C in Figure 4 have previously been assigned to the C2H imidazole-ring protons of the three histidines His- 18, His- 151, and His-2 1, respectively (Turner et al., 1983). The C2H signal of His-15 1mainly follows a simple one-proton titration, with values of PKa and chemical shifts virtually identical to the values found in small peptides and random-coil proteins (Markley, 1973; Bundi & Wiithrich, 1979). At 305 K a small, secondary shift with an apparent PKa value of 8.1 is observed. The C2H resonances of His-18 and His-21 show a more complex pH dependence. Apart from the major shift, caused by direct ionization of the imidazole ring, both signals are influenced by one or two more titrations (Tables I11 and IV). The titration of the imidazole ring of His-18, giving the major change in the chemical shift of the C2H proton signal, has an abnormally high PKa value (7.686 at 293 K, and 8.16 at 305 K), in agreement with the previous suggestion (Turner et al., 1983) that this residue is involved in an ion-pair interaction with a deprotonated carboxyl group. The value of the additional apparent titration, observed at acidic pH, shows that the interacting acidic group is likely to be an Asp or Glu side-chain carboxylate group with a correspondingly low PKa value (3.39 at 293 K). For the primary titration of His-18, a significant improvement in the overall RMS deviation at 293 K was obtained using a variable Hill coefficient for the titration characterized by pKa3. The fitted Hill coefficient of 2.55 (0.16) indicatesthat the titration of His- 18 is not a simple one-proton event. However, at 305 K the Hill coefficient did not differ significantly from unity. A minor shift with an apparent PKa value of 6.59 is observed for the His-18 C2H signal at 293 K. This value is close to the PKa value for the

8592 Biochemistry, Vol. 31, No.36, 1992

Abildgaard et al.

a

b

1' 1

L

PPm

I

I

I

I

I

8

6

4

2

0

FIGURE 4: 500 MHz 1H NMR sDectra at 305 K of human -growth hormone dissolved in D20 at (a) pD 8.5 and (b) pD 2.1. The signal marked

X is from dioxane.

I

I

1

I

I

I

I

I

8.6

8.4 8.2 8.0 7.8

7.6

t

6.4

1

60.8. i

0.6

0.4 0.2

4

2 4 6 8 PH FIGURE 5: Dependence on pH at 293 K of the 'H signals indicated in Figure 4. The solid lines correspond to the parameters given in Table 111. O

at 305 K His-21 shows a titration with an apparent PKa value (2.8) identical to the apparent, acidic pKavalueof His-18 and with a titration shift (approximately 0.2 ppm) that is even larger than for His- 18; that is, the titration of the acidic group that interacts with His-18 has an even larger effect on the chemical shift of His-21. Again, this indicates that an interaction is taking place between the two histidine residues. The low-field aromatic resonance D (7.63 ppm at pD 2.1, Figure 5) has an intensity of only one proton and shows a doublet fine structure at resolution enhancement. In the TOCSY spectrum recorded at pH 3.6 at 293 K (not shown), a connectivity with a proton at 6.92 ppm is observed. Thus, a possible assignment for resonance D is the C4H or the C7H ring proton of the only Trp residue in position 86. Other possibilities are the C2H or C6H of a rigid (not flipping) phenylalanine or one of the ring protons of a rigid tyrosine ring. However, no other Jconnectivities are observed for the signal scalar coupled to D, which might imply that D is belonging to a rigid tyrosine ring. At pH titration the signal follows a titration curve with a pKa of 4.00 and 4.27 at 293 and 305 K, respectively, leading to an upfield shift of 0.1 ppm at low pH. Resonance E (6.47ppm at pH 3.6,293K) with an intensity corresponding to two protons was assigned to the C3,5H ring protons of a Tyr residue, with the C2,6H proton at 6.62 ppm. The chemical shift shows two pH-dependent perturbations; a primary upfield shift on deprotonation with a PKa of 3.58 and a secondary transition with a PKa of 6.42 (293 K). In the NOESY spectrum of hGH in H20, an NOE connection is seen between signals D and E. Signal E is also showing an NOE to the C2,6H protons of an unnamed phenylalanine residue with the chemical shift of C2,6H, C3,5H, and C4H at 7.14, 6.80, and 7.40 ppm, respectively. The proton in E that is coupled to D also has an NOE to the C4H proton of the unnamed phenylalanine residue, indicating a close contact between the side chains of the two tyrosines, D and E,and the

primary titration of His-21 (6.62 at 293 K), indicating an interaction between these two histidine residues. Similarly,

Biochemistry, Vol. 31, No. 36, 1992 8593

Tertiary Interactions in Human Growth Hormone

Table 111: Ionization Constants and Chemical Shifts' for the IH Resonances of hGH at 293 K (PP4 PKa 1 62 ( P P d pKa2 63 (PPm) ~Ka3 64 ( P P d 8.508 6.59 8.346 7.686 7.643 8.607 3.39 (0.004) (0.08) (0.11) (0.019) (0.018) (0.002) (0.003) 6.779 7.681 His-151, C2H 8.6039 BC (0.0016) (0.008) (0.002) 6.623 7.407 8.286 His-21, C2H 8.457 2.69 C (0.019) (0.005) (0.016) (0.14) (0.007) 4.00 7.6348 7.7518 D Tyr, C2,6H (0.0013) (0.03) (0.0007) 6.42 6.3690 6.5832 3.580 6.3921 Tyr, C3,5H E (0.1 1) (0.0007) (0.0015) (0.018) (0.0017) 6.076 F Tyr, C33H (0.010) 3.91 0.4402 0.691 G Leu, CH3 (0.02) (0.0010) (0.002) 0.2780 0.5413 3.487 H Val/Leu, CH3 (0.0010) (0.0005) (0.015) 0.388 3.72 0.6356 0.427 2.65 Ile, CH3 I (0.003) (0.05) (0.008) (0.04) (0.0007) 4.68 0.1625 0.5825 J Val, CH3 (0.0019) (0.0012) (0.03) 0.198 Leu, CtlH3 K (0.008) - 0.065 L Leu, CT2H3 (0.01 1) a Parameters were obtained by nonlinear least-squares fit of eqs 6 and 7. Figures in parentheses are the calculated standard error derived from the diagonal elements of the variance-covariance matrix. Parameters obtained with a fitted Hill coefficient of 2.55 (0.16) for pKa3. Parameters obtained with a fitted Hill coefficient of 1.29 (0.02) for pKal. Ab

signal His-18, C2H

61

Table IV: Ionization Constants and Chemical Shifts' for the 'H Resonances of hGH at 305 K A

signal His-18, C2H

B

His-151, C2H

C

His-21, C2H

D

Tyr, C2,6H

E

Tyr, C3,5H

F

Tyr, C3,5H

G

Leu, CH3

H

Val/Leu, CH3

I

Ile, CH3

J

Val, CH3

K

Leu, CTlH3

L

Leu, CT2H3

(PPd 8.625 (0.015) 8.593 (0.004) 8.472 (0.008) 7.6268 (0.0012) 6.600 (0.006) 6.071 (0.012) 0.651 (0.002) 0.568 (0.002) 0.380 (0.007) 0.1985 (0.0008) 0.201 (0.016) - 0.02 (0.02) 61

PKa I 2.8 (0.4) 6.83 (0.04) 2.78 (0.07) 4.50 (0.05) 3.98 (0.07)

62

(PPm) 8.54 (0.04) 7.72 (0.04) 8.260 (0.006) 7.7270 (0.0008) 6.430 (0.007)

~Ka2 4.8 (0.5) 8.1 (0.3) 6.65 (0.03)

63 (ppm)

6.4 (0.4)

6.4033 (0.0016)

3.3 (0.3) 3.703 (0.0 19) 2.7 (0.5) 5.20 (0.04)

0.71 (0.04) 0.281 1 (0.0010) 0.42 (0.02) 0.531 (0.012)

4.16 (0.12)

0.508 (0.005)

4.01 (0.07) 7 .O (0.5)

0.6358 (0.0012) 0.549 (0.001)

8.412 (0.009) 7.627 (0.006) 7.477 (0.019)

~Ka3 8.16 (0.03)

(PPm) 7.551 (0.01 1)

64

8.3 (0.3)

7.405 (0.007)

6.3 (0.3)

0.483 (0.001)

a Parameters were obtained by nonlinear least-squares fit of eqs 6 and 7. Figures in parentheses are the calculated standard error derived from the diagonal elements of the variance-covariance matrix.

unnamed phenylalanine. The chemical shift of resonance F (6.09 ppm at pH 3.6), with an intensity of two protons, is almost unaffected by the changes in pH. On the basis of the doublet fine structure and a single connectivityto a resonance at 6.73 ppm in the TOCSY spectrum, it can be assigned to the upfield-shifted signal from the C3,5H ring protons of a Tyr residue. The six resonances G-L (Figure 4), all with intensities corresponding to three or six protons (vide infra), can be assigned to methyl groups with chemical shifts moved upfield relative to the corresponding signals in small peptides. The upfieldshifted position can be ascribed to the proximity of the methyl groups to aromatic side chains, and as such, the shift of these

resonances is very sensitive to conformational changes. A doublet fine structure of the three-proton resonance G~makes it assignableto the CH3 group of a Valor Leu residue. Further, in the COSY spectrum at pD 3.3 a connectivity to a signal at 2.04 ppm, which again correlates to a signal at 2.43 ppm, suggests that it is a Leu. A major pH-induced change in chemical shift is observed with a pK, value of 3.91 (293 K). In addition, two minor apparent titrations are observed at 305 K. The corresponding pKa values of 3.3 and 6.3 are relatively close to the values found for His- 18 and His-21, indicating a relation to these residues. Resonance H has an intensity corresponding to six protons and correlates to two resonances at 1.78 and 1.82 ppm, respectively; Le., the two methyl groups

8594 Biochemistry, Vol. 31, No. 36, 1992

Abildgaard et al.

FIGURE 6: Stereoview of the hGH molecule illustrating the helix-helix contact area between helix 1 and helix 4 (see text). The interactions between the aromatic side chains of the two helices are indicated, as well as the close proximity of His-18 and Asp-17 1 that allows the formation of a salt bridge between these two residues, the position of His-21, and the close proximity of Phe-31 and Leu-156 that gives rise to the extreme chemical shift of the CH3 protons of the leucine. For further details see text. in the H resonance correlate to different protons and must, therefore, belong to different residues. While the resonance at 1.78 is correlated to a signal in the CaH region (3.88 ppm) and, therefore, can be assigned to a CH3 group of a Val, the correlation of the resonance at 1.82 is unclear. Even so, H shows only one single and well-defined titration shift with a pKa value of 3.71, indicating that the two CH3 groups of the resonance are in adjacent locations. Resonance I also correlates to two resonances (0.58 and 1.07 ppm, respectively, at pD 3.3) but has the intensity of only three protons. It was, therefore, assigned to an Ile. On protonation it moves upfield, with the major titration shift corresponding to a PKa valueof 3.72 (293 K). An additional pH-induced shift of this resonance with a pKa value of 2.65 (293 K) is also observed. The three-proton resonance J and a resonance at 0.78 ppm (pD 3.3) are both correlated to a signal at 1.66 ppm, which, in turn, is correlated to a signal in the a-region at 4.18 ppm. Resonance J is therefore assigned to a Val. At protonation, resonance J is upfield shifted following a simple ionization curve with a pKa value of 4.68 (293 K). Its line width is markedly greater at the midpoint of the titration. The broadening is attributed to exchange effects associated with the structural interconversion caused by the ionization process (vide infra). From the exchange broadening the exchange rate was estimated to kA = 300 s-1. The methyl groups K and L both correlate to the same 0proton. They are, therefore, assigned to the CY1H3 and CYzH3 groups of the same Val residue. They show no significant pH dependence.

DISCUSSION In globular proteins the dispersion of the backbone carbonyl carbon resonances, corresponding to a given type of amino acid residue, is due mainly to the differences in local conformation, as defined by the backbone dihedral angles and the manner of hydrogen bonding (Sait6, 1986; Ando et al., 1988). In random coil proteins the backbone carbonyl carbons of the same type of residue are effectivelyindependent of the neighboring residues (Howarth & Lilley, 1978), and for leucine residues in proteins in aqueous solution the random coil chemical shift value of the 13Ccarbonyl carbon is 175.6 ppm (Howarth & Lilley, 1978). Consequently, the large dispersion of 5 ppm of the resonances from the 26 hGH leucine residues, observed in the entire pH region in Figure 1,

shows that although numerous pH-induced conformational changes occur, these changes do not lead to any substantial unfolding of the protein. These findings are in accordance with the results from optical rotatory dispersion measurements (Bewley & Li, 1967) and CD spectra (Aloj & Edelhoch, 1972), which show that the secondary structure of hGH is virtually unaffected by changes in pH. The three-dimensional structure of hGH is presently not known in detail, and our knowledge of the folding is based on secondary structure predictions (Chou & Fasman, 1974a,b; Kawauchi & Li, 1974) and the results derived from various physicochemical methods. Recently, the crystal structure at 2.8-A resolution of the highly homologous porcine growth hormone (pGH) has been reported in some detail (AbdelMeguid et al., 1987). Due to the extensive degree of amino acid sequence homology with human growth hormone, particularly within the four helical regions known for pGH, it is reasonable to assume that the three-dimensional structure of hGH is similar to that of pGH. Given the validity of this assumption, the structure of hGH is, basically, composed of a tightly packed bundle of four antiparallel a-helices made up of residues 6-33 (helix l ) , 7596 (helix 2), 106-129 (helix 3), and 154-183 (helix 4). Although the atomic coordinates for pGH are presently not available, some of the general features of the helix-helix packing can be established. Thus, helix 1 and helix 4, both 45 A in length, form one-third of a turn of an antiparallel coiled-coil a-helix (Crick, 1953; Schulz & Schirmer, 1979). This helix-helix packing is especially favored by side-chain meshing of hydrophobic residues at the central positions of the contact area (i, i + 3, and i‘, i’- 4, where i = j + 7k and i’= j ’ - 7k while k = 1,2, ...), by polar residues at the outer surface positions (i 1, i 2, i 5, and i’- 2, i’- 5, i’- 6), and by charged residuesin favorable positions for the formation of salt bridges or hydrogen bonds with their counter parts ( i 4, i 6, and i‘- 3, i‘- 8, respectively). Letting j = 7 and j ’ = 181, the resulting helix-helix packing enables the formation of a salt bridge between His-18and Asp-171, as suggested by the abnormal pKa value of His-18 and its significant apparent titration at acidic pH. A model of hGH showing this interaction is sketched in Figure 6. The structure shown in Figure 6 was derived from the coordinates of the calculated structure of bovine growth hormone (Carlacci et

+

+

+

+

+

Tertiary Interactions in Human Growth Hormone al., 1991) using the program HOMOLOGY (Biosym Technologies, Inc.). HOMOLOGY facilitates creating a three-dimensional structure for a protein from a known amino acid sequence and a known structure for a related protein. This initial structure was subjected to alternating restrained energy minimization and restrained molecular dynamics (Kaptein et al., 1985) using the program DISCOVER (Biosym Technologies, Inc.). With the interhelical packing scheme shown in Figure 6, the His-21 resides at the helix-helix contact area inaccessible to the solvent. This position explains the low exchange rate constant of the His-21 imidazole C2H proton previously observed (Turner et al., 1983). Moreover, the spatial arrangement of the two helices will allow for favorable interactions between the aromatic side chains (Serrano et al., 1991) of the residues His- 18, His-21, Phe- 166, Phe-25, Tyr28, Tyr-160, and Phe-3 1, as indicated in Figure 6. Thus, the close contact between His-21 and Phe-166 and Phe-25 is probably responsible for the upfield-shifted position of the His-21 C2H proton resonance (resonance C, Figure 5 ) . The three interacting aromatic side chains (D, E, and the unnamed phenylalanine) may be part of this set of aromatic side chains. The fact that resonance E is affected by the titrations of the side chain of His-21 supports this suggestion. A tentative assignment would then be: E and D are Tyr-28 and Tyr-160, in some order, and the unnamed phenylalanine is either Phe31 or Phe-25. Finally, the significant change of the shielding of the His-21 C2H proton by titration of the acidic group that interacts with His-18 suggests that the change in ion-pair interaction causes a change in the relative position of helix 1 and helix 4. This change alters the relative position of the stacked aromatic side chains, leading to the observed changes in chemical shift for these resonances. Unlike His- 151 and His-2 1, the pKa value of His- 18 shows a profound temperature dependence (an increase of 0.5 going from 293 to 305 K). Therefore, this pKa value provides information about the ion-pair interaction. Qualitatively, the temperature variation shows that the deprotonation of the imidazole cation decreases with increasing temperature, indicating that the stability of the salt bridge increases with the temperature. Quantitatively, the difference between the two PKa values corresponds to a AH = -68 kJ mol-' for the deprotonation of the imidazolegroup of His- 18. This contrasts with the AH value for the deprotonation of a noninteracting imidazole group in a protein which is positive and on the order of 30 kJ mol-' (Roberts et al., 1969; Fersht, 1985). Therefore, the AHvalue obtained here indicates that the salt bridge formation is opposed by an increase in enthalpy of about 95 kJ mol-'. This clearly shows that the salt bridge formation is entropy driven (Cantor & Schimmel, 1980b). Further, the change of the dissociation constant, caused by the ionic interaction in the salt bridge, is given by the free energy change, AGc, which is related to the pKa values by the equation (Cantor & Schimmel, 1980a,b)

Here pKL is the observed pKa value and pKa,0is the pKa value in the absence of the electrostatic interaction. With a pKa,o of a noninteracting histidine of 6.8 1 at 305 K (Markley, 1973), the p K l value of 8.16 found here corresponds to a free energy of stabilization of the salt bridge of approximately -7.2 kJ mol-' at 305 K, according to eq 10. Assuming that the free energy change responsible for the lowering of the pKa (2.8 at 305 K) of the carboxylic acid of the salt bridge is the same, we find that the pKa,0 in the absence of the interaction is

Biochemistry, Vol. 31, No. 36, 1992 8595 approximately 4.1. This is consistent with a side-chain car boxyl group of an Asp residue (Nozaki & Tanford, 1967; Cantor & Schimmel, 1980a) and, thus, with the proposition that the interacting group is Asp-171 (vide supra). The close resemblance between the weighted average of pKal for His-18 and His-21 at 293 K and the weighted average pKa of the leucine carbonyls in group I (Table 11), suggests that the structural changes associated with the leucine carbonyl carbons in group I are linked to the titration of the Asp- 171 side-chain group. The fact that only four leucine carbonyl resonances are affected significantly by the change in the His-18 ion-pair interaction suggests that the conformational transition, associated with the formation and breaking of the salt bridge between His-18 and Asp-171, is confined to the region around His-18 and His-21. A deformation of the two helices in this region (helix 1 and helix 4), accompanied by a change in the hydrogen bonding of the secondary structure and the main-chain dihedral angles, might lead to the observed changes of the chemical shifts of the carbonyl carbons (Ando et al., 1984; Sait6, 1986). As indicated by the resonances 1 and 4 (Table I), the interconversion between the two forms is relatively fast ( k =~ 770 s-l, T A = 1.3 ms). Theincrease in the strength of the electrostatic interactions with temperature is probably responsible for the greater number of pH-induced shifts observed for the proton resonances at 305 K. With the increased strength of these interactions, their disruption is expected to have a more farreaching effect on the structure and, thus, on the chemical shift of the proton signals. Fluorescence emission spectroscopy of hGH (Aloj & Edelhoch, 1972) has shown that the pH dependence of the quenching of the tryptophanyl fluorescence from the single tryptophan residue in position 86 fits a simple ionization process with a PKa of 5.0 at 298 K. The only proton signal with an apparent titration in this range is resonance J, assigned to an upfield-shiftedCH3 group of a Val residue. This residue shows an apparent titration shift with a pKa of 4.68 at 293 K and 5.20 at 305 K, which, by interpolation, gives a predicted pKa of 4.9 at 298 K. The pGH crystal structure shows that Trp86 resides in helix 2. This helix, in turn, is packed in the interior of the molecule between the coiled-coil a-helix (helix 1 and helix 4), helix 3, and the irregular peptide fragment connecting helix 3 with helix 4. Assuming the same a-helical structure around Trp-86 in hGH, the Val-90 residue is close to Trp-86. This position may result in a considerably upfield shift of the Val-90 CH3 group due to a ring current shift from Trp-86. Therefore, both the pKa of the apparent titration of the Val CH3 resonance, J , and its extreme chemical shift suggest that this resonance can be assigned to one of the CH3 groups of Val-90. The apparent pK, of 4.8 for the Leu carbonyl resonances forming group VI1 implies that this group is affected by the protonation event responsible for the change in the tryptophanyl fluoresence emission and the concomitant apparent 'H titration shift of Val-90. The observations just mentioned seem to indicate that localized conformational changes accompany the titration of a single acidic group around pH 5 . A local deformation of helix 2, leading to a change in the main-chain dihedral angles, can explain the change in the position of the Val-90 CH3 group with respect to Trp-86, as well as the change in chemical shift of the four leucine carbonyl carbons forming group VII. The three leucines close to Trp-86 (Leu-80, Leu-8 1, and Leu82) may be members of group VII. An ionization process with a pKa value of this size can be ascribed to the deprotonation of an Asp or a Glu side-chain group that either

8596 Biochemistry, Vol. 31, No. 36, 1992 interacts with a nearby negatively charged group or is buried inside a hydrophobic environment of the protein. This makes Glu- 174 a possible candidate for the acidic group. This residue is probably buried in the interior of the helix 1, helix 4, helix 2 contact area where it may be partly stabilized by one or more of the nearby polar groups Met-14, Met-170, and Arg178. Thus, the exchange rate constant, k A = 370 s-I ( T A = 2.7 ms), observed for the leucine carbonyl carbons in group VII, as well as for the Val-90 CH3 resonance (=300 s-I, vide supra) may reflect the rate of rupture of the local structure necessary to make the Glu-174accessible to the solvent. The upfield-shifted methyl resonances G-I all show apparent titrations around pH 4. Several Leu carbonyl resonances also titrate in this region (Figure 3). A possible assignment to signal G is the leucine residue in position 156. As shown in Figure 6, Leu- 156located near Phe-3 1could be shifted upfield due to the interaction with the aromatic side chain. A minor pH-induced shift with an apparent pKa of 6.3, observed for signal G, seems to indicate that the local structure around Leu-156 is slightly dependent on the titration of His-21. At least two additional structural transitions are seen. One at an pKa of 3.8 might involve the resonances G and I since both have a major apparent titration at the same pKa. However, it is difficult to draw any definite conclusions about the assignment of the signals based upon the information obtained so far. CONCLUSIONS

The present study demonstrates that the backbone carbonyl 13CNMR resonanas are sensitive probes of the local structure of folded proteins, and the investigation here of hGH shows that the combination of specificisotope labeling and a complete least-squares analysis allows the detection of even small changes in the globular structure of a relatively large protein. In combination with titration shifts observed in the proton spectrum, the study here also shows that the pH-induced conformational changes in hGH can be described as a series of localized adjustments of the tertiary structure imposed by the change in favorable interactions between different residues in the globular folded protein. No indications of a profound unfolding of the globular structure were observed. The conformational changes associated with the deprotonation of acidic side-chain groups involved in ion-pair interactions are relatively fast, with preexchange lifetimes on the order of 1 ms or less. The structural changes associated with the deprotonation of the buried Glu-174side chain is somewhat slower, presumably because a more drastic perturbation of the folded structure is required to permit water to penetrate into the interior of the protein. ACKNOWLEDGMENT

We thank Drs. Henrik Gesmar and Georg Ole Sarensen for helpful discussions and Mrs. Else Philipp for recording some of the proton spectra. REFERENCES

Abdel-Meguid,S. S., Shieh, HA., Smith, W. W., Dayringer, H. E., Violand, B. N., & Bentle, L. A. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 6434-6437. Abildgaard, F., Gesmar, H., & Led, J. J. (1988) J.Magn. Reson. 79, 78-89.

Abildgaard, F., Sarensen, G. O., & Led, J. J. (1991) J. Magn. Reson. 91, 148-154. Aloj, S . , & Edelhoch, H. (1972) J. Biol. Chem. 247, 1146-1 152. Ando, I., Tabeta, R., Shoji, A,, & Ozaki, T. (1984) Macromolecules 17, 457-461.

Abildgaard et al. Ando, S., Ando, I., Shoji, A., & Ozaki, T. (1988) J . Am. Chem. SOC.110, 3380-3386. Bewley, T. A., & Li, C. H. (1967) Biochim. Biophys. Acta 140, 201-207.

Bewley, T. A., & Li, C. H. (1975) in Aduances in Enzymology (Meister, A., Ed.) Vol. 42, p 87, Wiley, New York. Bundi, A., & Wuthrich, K. (1979) Biopolymers 18, 285-297. Burger, H. G., Edelhoch, H., & Condliffe, P. G. (1966) J . Biol. Chem. 241,449-457. Cantor, C. R., & Schimmel,P. R. (1 980a) Biophysical Chemistry pp 42-5 1, Freeman, San Francisco. Cantor, C. R., & Schimmel,P. R. (1980b) BiophysicalChemistry, pp 289-291, Freeman, San Francisco. Carlacci, L., Chou, K.-C., & Maggiora, G. M. (1991) Biochemistry 30, 4389-4398. Chou, P. Y., & Fasman, G. D. (1974a) Biochemistry 13, 211222.

Chou, P. Y., & Fasman, G. D. (1974b) Biochemistry 13, 222244.

Christensen,T., Hansen, J. W., Pedersen,J., Dalbsge, H., Carlsen, S., Jensen, E. B., Jsrgensen, K. D., Dinesen, B., Nilsson, P., Ssrensen, H. H., Thomsen, J., & Kappelgaard, A.-M. (1986) NATO ASZ Ser., Ser. A 125, 305-3 15. Christensen, T., Jensen, E. B., Junker, F., Dalbsge, H., Abildgaard, F., & Led, J. J. (1992) Acta Chem. Scand. 46,97-99. Crick, F. H. C. (1953) Acta Crystallogr. 6, 689-697. Dalbage, H., Dahl, H.-H. M., Pedersen, J., Hansen, J. W., & Christensen, T. (1987) BiolTechnology 5, 161-164. Eigen, M. (1963) Angew. Chem. 75,489-508. Eigen, M., Hammes, G. G., & Kustin, K. (1960) J. Am. Chem. SOC.82, 3482-3483. Fersht, A. (1985) Enzyme Structure and Mechanism, 2nd ed., p 173, Freeman, New York. Golub, G. H., & van Loan, C. F. (1980) SZAM J. Numer. Anal. 17, 883-893.

Holladay, L. A., Hammonds, R. G., & Puett, D. (1974) Biochemistry 13, 1653-1661. Holzman, T. F., Dougherty, J. J., Brems, D. N., & MacKenzie, N. E. (1990) Biochemistry 29, 1255-1261. Howarth, 0.W.,& Lilley, D. M. J. ( 1978) Prog. NMR Spectrosc. 12, 1-40.

Hughson, F. M., & Baldwin, R. L. (1989) Biochemistry 28,441 54422.

Kalinichenko, P. (1976) Stud. Biophys. 58, 235-240. Kaplan, J. I., & Fraenkel, G. (1980) NMR of Chemically Exchanging Systems, pp 74-80, Academic Press, New York. Kaptein, R., Zuiderweg, E. R. P., Scheek, R. M., Boelens, R., & van Gunsteren, W. F. (1985) J. Mol. Biol. 182, 179-182. Kawauchi, H., & Li, C. H. (1974) Arch. Biochem. Biophys. 165, 255-262.

Kim, P. S., & Baldwin, R. L. (1982) Annu. Rev. Biochem. 51, 459-489.

Markley, J. L. (1973) Biochemistry 12, 2245-2249. Nall, B. T., Osterhout, J. J., Jr., & Ramdas, L. (1988) Biochemistry 27, 73 10-73 14. Nozaki, Y., & Tanford, C. (1967) J . Biol. Chem. 242, 47314735.

Roberts, G. C. K., Meadows, D. H., & Jardetzky, 0. (1969) Biochemistry 8, 2053-2056. Russell, A. J., & Fersht, A. R. (1987) Nature 328, 496-500. Sait6, H. (1986) Magn. Reson. Chem. 24, 835-852. Schultz, G. E., & Schirmer, R. H. (1979) Principles of Protein Structure, pp 79-8 1, Springer-Verlag, New York. Serrano, L., Bycroft, M., & Fersht, A. R. (1991) J. Mol. Biol. 218,465-475.

Shaka, A. J., Keeler, J., Frenkiel, T., & Freeman, R. (1983) J. Magn. Reson. 52, 335-338. Tiichsen, E., & Hansen, P. E. (1988) Biochemistry 27, 85688576.

Turner, C., Cary, P. D., Grego,B., Hearn, M. T. W., & Chapman, G. E. (1983) Biochem. J. 213, 107-113. Registry No. GH,9002-72-6; H+,12408-02-5.